Abstract
Registration of liver US–MR data is essential for the multimodal assessment of liver diseases, yet accurate alignment remains challenging due to the differences in contrast profiles between modalities. In this work, we present a multi-task learning framework for robust liver US–MR image alignment. The proposed method jointly predicts anatomical segmentation maps for a US–MR image pair and an initial rigid transformation, and then performs rigid- and non-rigid instance optimization using image features, segmentation, and surface models to refine the alignment. The proposed method was validated on our in vivo liver dataset and achieved an average TRE of 4.29±1.09 mm, while maintaining reliable performance under perturbations up to ±30 mm in translation and ±45◦ in rotation, demonstrating improved performance and robustness compared with existing methods.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_015.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=xP6E6fY0Jx
BibTex
@InProceedings{ZenQi_Spatially_MICCAISAT2026,
author = { Zeng, Qi AND Kemper, Tara AND Li, Zongze AND Chen, Wanwen AND Pang, Emily H. T. AND Rohling, Robert AND Salcudean, Septimiu E.},
title = { { Spatially Invariant Multi-Task Learning for Liver Ultrasound–MRI Registration } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17272},
month = {pending},
page = {pending}
}